influence of plankton community structure on the sinking ... · the enclosed plankton community...

Influence of plankton community structure on the sinkingvelocity of marine aggregatesL. T. Bach1, T. Boxhammer1, A. Larsen2, N. Hildebrandt3, K. G. Schulz1,4, and U. Riebesell1

1GEOMAR Helmholtz Centre for Ocean Research Kiel, Kiel, Germany, 2Hjort Centre for Marine Ecosystem Dynamics, UniResearch Environment, Bergen, Norway, 3Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research,Bremerhaven, Germany, 4Centre for Coastal Biogeochemistry, Southern Cross University, East Lismore, New South Wales,Australia

Abstract About 50 Gt of carbon is fixed photosynthetically by surface ocean phytoplankton communitiesevery year. Part of this organic matter is reprocessed within the plankton community to form aggregateswhich eventually sink and export carbon into the deep ocean. The fraction of organic matter leaving thesurface ocean is partly dependent on aggregate sinking velocity which accelerates with increasing aggregatesize and density, where the latter is controlled by ballast load and aggregate porosity. In May 2011, wemoored nine 25m deep mesocosms in a Norwegian fjord to assess on a daily basis how plankton communitystructure affects material properties and sinking velocities of aggregates (Ø 80–400μm) collected in themesocosms’ sediment traps. We noted that sinking velocity was not necessarily accelerated by opal ballastduring diatom blooms, which could be due to relatively high porosity of these rather fresh aggregates.Furthermore, estimated aggregate porosity (Pestimated) decreased as the picoautotroph (0.2–2μm) fraction ofthe phytoplankton biomass increased. Thus, picoautotroph-dominated communities may be indicative forfood webs promoting a high degree of aggregate repackaging with potential for accelerated sinking. Bloomsof the coccolithophore Emiliania huxleyi revealed that cell concentrations of ~1500 cells/mL acceleratesinking by about 35–40%, which we estimate (by one-dimensional modeling) to elevate organic mattertransfer efficiency through the mesopelagic from 14 to 24%. Our results indicate that sinking velocities areinfluenced by the complex interplay between the availability of ballast minerals and aggregate packaging;both of which are controlled by plankton community structure.

1. Introduction

Phytoplankton fixes approximately 50 Gt of carbon in the euphotic zone of the oceans every year [Longhurstet al., 1995; Field et al., 1998]. Most of this organic carbon is remineralized in the surface ocean. However,5–12Gt C escape remineralization and are exported out of the surface (100m), either as dissolved organiccarbon (DOC) trapped in downwelling water currents, as particulate organic carbon (POC) sinking due togravity, or transported downward by vertically migrating zooplankton [Hansell and Carlson, 2001; Turner,2002; Honjo et al., 2008; Steinberg et al., 2008; Henson et al., 2011; Siegel et al., 2014]. A large fraction of thePOC and DOC is biologically remineralized during its descent through subsurface water and released asdissolved inorganic carbon (DIC). That way, DIC is transported from surface to depth against a concentrationgradient in a series of biologically mediated processes—the reason why this has been termed the biologicalpump [Volk and Hoffert, 1985]. A shutdown of the biological pump would lead to a significant accumulationof DIC in the surface ocean paralleled by a roughly 60–70% increase of current atmospheric CO2 concentra-tions within a 1000 year equilibration time [Maier-Reimer et al., 1996]. This highlights the outstanding impor-tance of the biological pump for the global carbon cycle and climate.

The fraction of surface-originating POC that reaches the deep ocean depends on the balance betweenparticle remineralization rates and sinking velocities. A fast-sinking particle made of relatively refractoryPOC will experience little remineralization during its descent from the euphotic zone (~0–200m) throughthe mesopelagic (200–1000m) into the deep ocean (below 1000m). Conversely, a slowly sinking particlemade of labile and easily disintegrating POC would never reach that far. Remineralization rates are controlledby bacterial activity and mesozooplankton feeding as well as particle fragmentation rates [Kiørboe, 2001;Giering et al., 2014]. Sinking velocity is determined by the viscosity of seawater, particle size, shape, and excessdensity [Stokes, 1850; McNown and Malaika, 1950; Smayda, 1970], where the latter depends primarily onporosity of the particle and the amount of ballast mineral attached to it [Alldredge and Gotschalk, 1988].

BACH ET AL. SINKING VELOCITY OF MARINE AGGREGATES 1145

PUBLICATIONSGlobal Biogeochemical Cycles

RESEARCH ARTICLE10.1002/2016GB005372

Key Points:• Aggregate sinking speeds arecontrolled by processes in theplankton community

• Plankton communities can affectsinking speed by altering aggregateporosity and ballasting

• Emiliania huxleyi blooms of1500 cells/mL increase organic mattertransfer efficiency from 14 to 24%

Supporting Information:• Supporting Information

Correspondence to:L. T. Bach,[email protected]

Citation:Bach, L. T., T. Boxhammer, A. Larsen,N. Hildebrandt, K. G. Schulz, andU. Riebesell (2016), Influence ofplankton community structure on thesinking velocity of marine aggregates,Global Biogeochem. Cycles, 30,1145–1165, doi:10.1002/2016GB005372.

Received 5 JAN 2016Accepted 6 JUL 2016Accepted article online 9 JUL 2016Published online 13 AUG 2016

©2016. American Geophysical Union.All Rights Reserved.

http://publications.agu.org/journals/

http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1944-9224

http://dx.doi.org/10.1002/2016GB005372

http://dx.doi.org/10.1002/2016GB005372

http://dx.doi.org/10.1002/2016GB005372

http://dx.doi.org/10.1002/2016GB005372

http://dx.doi.org/10.1002/2016GB005372

mailto:[email protected]

Opal, CaCO3, and lithogenic minerals are ballast materials formed in or deposited onto the oceans’ surfacelayer. Opal is formed by diatoms, silicoflagellates, and radiolarians, while CaCO3 is mainly formed by cocco-lithophores, calcifying dinoflagellates, foraminifera, and pteropods. Lithogenic minerals are usually of terrige-nous origin and transported into the oceans via dust events or river input. Ballast minerals are importantcomponents in particle export as they may slow down remineralization of associated POC [Hedges andOades, 1997] and increase particle excess density, thereby accelerating sinking velocity [e.g., Passow andDe La Rocha, 2006; Honjo et al., 2008; Ploug et al., 2008]. These mechanisms were used to explain the ballastratio hypothesis [Armstrong et al., 2009], which originates from the observation that globally averaged POC:ballast mineral ratios are high and variable in the euphotic zone but converge to relatively narrow and stablevalues in the deep [Armstrong et al., 2002]. The ballast ratio hypothesis suggests that ballast minerals, in par-ticular relatively dense CaCO3, strongly support deep ocean POC sequestration since most organic materialreaching below 2000m is associated with them [Francois et al., 2002; Klaas and Archer, 2002].

Recently, however, the ballast ratio hypothesis has been called into question. Passow and De La Rocha[2006] and Boyd and Trull [2007] noted that globally averaged correlations between ballast materials andPOC may not be found on regional scales or along the course of an entire seasonal cycle. This concernwas confirmed by the analysis of Wilson et al. [2012] and Le Moigne et al. [2014], who found variable regio-nal correlations between CaCO3 and POC in the deep and in the surface ocean and consequently attributedthe tight global correlation as an artifact of spatial averaging. Francois et al. [2002], Lam and Bishop [2007],Lam et al. [2011], and Henson et al. [2012a, 2012b] reported that ballast-rich diatom blooms are character-ized by surprisingly low POC transfer efficiencies into the deep ocean and explained this finding by (1) thehigh degree of fluffiness of diatom aggregates and (2) the relatively large proportion of easily degradableorganic carbon compounds within diatom aggregates. In consequence, they suggested that POC transferefficiency is not primarily determined by the presence of ballast minerals but by the upper ocean ecosys-tem structure with those systems producing tightly packed and refractory aggregates having highest trans-fer efficiencies. It is important to keep in mind, however, that these recent developments do not exclude animportant influence of ballast under all circumstances. Instead, they shift the focus from ballast materials asprimary controlling factor of export toward the state of the pelagic ecosystems being of major relevance forexport flux. Nevertheless, ballast materials may seasonally and/or regionally still be very important [e.g.,Waite et al., 2005; Honda and Watanabe, 2010; Martin et al., 2011; Smetacek et al., 2012].

In this study we used mesocosms to follow the development of particle sinking velocities over time and con-nect it to the succession of natural plankton communities in order to evaluate the influence of food webprocesses on sinking.

2. Materials and Methods2.1. Experimental Setup

In May 2011, nine “Kiel Off-Shore Mesocosms for future Ocean Simulations” (M1–M9) were deployed for5weeks in a Norwegian fjord (Raunefjord; 60.265°N, 5.205°E) close to the city of Bergen. The cylindrical butinitially folded mesocosm bags made of transparent polyurethane foil were mounted in 8m long floatingframes [Riebesell et al., 2013]. After mooring of the flotation frames at the study site, bags were unfoldedby lowering the bottom part to 24m, thereby enclosing the natural plankton community. Bottom and topof the cylindrical bags were covered with 3mm meshes before extension to exclude patchily distributedlarger plankton (e.g., adult jelly fish) and nekton (e.g., fish). The mesh-covered mesocosms were allowed toexchange with seawater for 3 days. After this period, scuba divers closed the mesocosms by removing thebottom mesh and then quickly sealing the wide opening with a 2m long conical sediment trap. The upperopenings of the bags were pulled above the water surface directly after trap attachment, thereby isolatingthe enclosed plankton community from the surrounding fjord water. The mesh covering the upper openingwas removed thereafter.

All mesocosms were 25m deep (Figure 1), 2m in diameter, and contained a volume of approximately 75m3.Sampling started after the mesocosms were closed, and the water column was mixed for 5min by bubblingwith compressed air eliminating a slight salinity stratification [Riebesell et al., 2013]. Salinity was ~32 through-out the entire water column after mixing and decreased to ~31.9 at the end of the experiment due to dilution

Global Biogeochemical Cycles 10.1002/2016GB005372


by rainwater. All mesocosms were left unper-turbed for the first 4 days of the experiment.Subsequently, seven of them were enrichedwith different amounts of CO2-saturated sea-water as explained by Riebesell et al. [2013] toreach initial pCO2 levels of 300, 310 (bothuntreated controls), 395, 590, 890, 1165,1425, 2060, and 3045μatm. The mesocosmwith 300μatm (M2) was discovered to havean unmendable hole and was thus excludedfrom analyses. All mesocosms were enrichedwith ~5μmol L�1 NO3

� and ~0.16μmol L�1

PO43� in the middle of the experiment

(Julian day 142; 22 May) as described bySchulz et al. [2013].

2.2. Sampling and Processing of WaterColumn Parameters

Phytoplankton, particulate matter, and dis-solved inorganic nutrient samples were takenevery morning between 09:00 and 11:00 A.M.with integrating water samplers (Hydro-Bios),which collect equal amounts of water fromevery depth (0–23m). Depth-integrated sam-ples were stored in 10 L carboys, transportedto land, and kept in the dark at in situ tempera-ture until subsampling for flow cytometry,chlorophyll a, biogenic silica (BSi), and inor-ganic nutrients. Great care was taken to mixthe carboys before every subsampling. Thetime between mesocosm sampling and sub-sampling from the carboys was usually wellbelow 3 h. Flow cytometry subsamples(50mL) were stored at in situ temperature fora maximum of 3 h until analysis with aFACSCalibur flow cytometer (BD Biosciences).Phytoplankton enumerations were obtainedfrom fresh samples based on difference inchlorophyll autofluorescence and side scatter

with the trigger set on red fluorescence [Larsen et al., 2001]. Light microscopy investigations revealed thathardly any phytoplankton cells >50μm were present in the samples, making the flow cytometry measure-ments (covering the size range <100μm) representative for the overall size spectrum. Subsamples for chlor-ophyll a and BSi were filtered with gentle vacuum (200mbar) on glass fiber filters or cellulose acetate filters,respectively. Both sample types were stored at �20°C until measurements following Welschmeyer [1994](chlorophyll a) or Hansen and Koroleff [1999] (BSi). Nitrate (NO3

�), phosphate (PO43�), and silicate (Si(OH)4)

concentrations were determined following Hansen and Koroleff [1999] using a Hitachi U2000 spectrophot-ometer [Schulz et al., 2013].

Zooplankton samples were collected with an Apstein net (55μmmesh size, 0.17m net opening) on a weeklybasis. The maximum sampling depth was 23m to prevent any contact of the net with the sedimenttrap. Sampling was restricted to four net hauls per mesocosm and week in order to avoid overfishing.Zooplankton were transported to the lab in less than an hour, where it was preserved withhexamethylenetetramine-buffered formalin (4% (v/v)) for counting and taxonomic analyses with astereomicroscope.

Figure 1. Schematic drawing of the KOSMOS system [Riebesellet al., 2013]. The video by Boxhammer et al. [2015] gives animpression on the plankton community enclosed in thesemesocosms.



2.3. Sampling and Processing of Sediment Trap Material

The bottom of the mesocosms had the conical shape of a sediment trap (Figure 1). Sinking material collectedin the traps was pumped daily between 08:00 and 09:00 A.M. through a 25m long (10mm inner diameter)silicon tube under low vacuum into a glass bottle at the surface [Boxhammer et al., 2016]. The bulk of thesediment trap material (>97%) was used to determine the amount and composition of particulate matter,while small subsamples were used for zooplankton counting and particle sinking velocity measurements(see below).

Bulk samples were concentrated by centrifugation. Resulting pellets were freeze-dried, weighed, and groundto fine powder as described by Boxhammer et al. [2016]. The powder was used for POC, total particulate car-bon (TPC), and BSi measurements. Before analysis, POC samples were soaked with 50μL of 1 molar HCl toremove all CaCO3. TPC and POC samples were subsequently analyzed on a C/N elemental analyzer(Hekatech). Particulate inorganic carbon (PIC) was calculated from the difference of TPC and POC. Theamount of sediment trap BSi was measured photometrically according to Hansen and Koroleff [1999].

Results from PIC, POC, and BSi analyses were used to estimate the relative contribution of each of theseconstituents to the material weight of the collected aggregates excluding water. Therefore, we assumed(1) material weights (weightmaterial) of 1.06, 2.1, and 2.7 g cm�3 for POM, BSi, and PIC, respectively[Sarmiento and Gruber, 2006] and (2) that no components other than these three contributed to the weightof the collected material. Assumption 2 is reasonable because there was no significant source of lithogenicballast materials inside a mesocosm, and all lithogenic particles present at the beginning of the experimentshould sink out during the first days (see section 4.2.1). Accordingly, the contribution of biogenic materials tothe total material weight (weightfraction) is calculated as

weightfraction ¼ b � weightmaterial

1:06 � POC þ 2:1 � BSi þ 2:7 � PIC(1)

where b is either POM, BSi, or PIC and weightmaterial is their respective densities.

Zooplankton samples (10mL) were transferred into glass vials and preserved with hexamethylenetetramine-buffered formalin (4% (v/v)). Samples from two consecutive days but from the same mesocosm were thenpooled and subsequently analyzed with a stereomicroscope.

Samples for sinking velocity measurements were carefully sieved (300μm) to exclude very large gelatinouszooplankton, which can (when present) clog the settling column [Bach et al., 2012a]. Samples were dilutedwith filtered (0.2μm) seawater (salinity ~33.7) to keep particle concentrations low enough to avoidparticle-particle interactions in the settling column, which are known to accelerate sinking velocities [Bachet al., 2012a]. This preparation procedure had little effect on fecal pellet integrity, and we also did not observethat pellets were retained on the sieve [Bach et al., 2012a]. Fluffy aggregates, however, did most likely notmaintain their original size during sampling preparation. Instead, they continuously disintegrated and reag-gregated so that their size, density, and porosity measured in the settling column are potentially differentthan they were when they were sinking in situ. The methodological uncertainties will be outlined insection 2.5, and limitations will be further discussed in section 4.2.2.

2.4. Particle Sinking Velocity and Size Measurements

Sinking velocities of the sediment trap samples weremeasured directly after preparation for ~20min with theFlowCam method described in Bach et al. [2012a]. This camera-based method allows parallel measurementsof sinking velocity, equivalent spherical diameter (ESD), and shape of each particle sinking through the set-tling chamber. Turbulence within the settling column induced by convection was suppressed by operatingthe system in a temperature-controlled room (10°C) and constantly ventilating the settling chamber [Bachet al., 2012a]. The size of the settling column is relatively small (400mm long× 10mm wide × 3mm deep)due to technical restrictions of the camera. This also limits the size of aggregates that can be reasonablyinvestigated to a maximum of 400μm, because significantly larger aggregates would experience too pro-nounced wall effects. Sinking aggregates were divided in four size classes according to their ESD to facilitatedata evaluation and discussion. Size classes were 80–130μm, 120–180μm, 170–260μm, and 240–400μm.The slight overlap between them was necessary to avoid exclusion of aggregates with an ESD exactly onthe border of two size classes [Bach et al., 2012a].



2.5. Aggregate Density and Porosity Calculations

Aggregate density (ρaggregate) was calculated from measured sinking velocity (Umeasured) and ESD usingStokes’ law:

ρaggregate ¼Umeasured�μseawater

29�g� ESD

2

� �2 þ ρseawater (2)

where g is the Earth’s gravitational acceleration, ESD is the equivalent spherical diameter of the aggregate,μseawater is the viscosity of seawater, and ρseawater is the density of seawater. The μseawater and ρseawater werecalculated from known salinity and temperature with the equations provided by Sharqawy et al. [2010].

The application of Stokes’ law for calculations of ρaggregate is appropriate when particles are equiaxial and sinkin a laminar flow regime. Laminar flowwas established in our measurements since investigated particles sanksufficiently slow and were relatively small so that their Reynolds numbers always remained far below thecritical threshold of 0.5 [McNown and Malaika, 1950]. Aggregate shape was random but the aspect ratio(i.e., length/width) of the majority of aggregates was above 0.8, also indicating a higher abundance of debrisrelative to fecal pellets. We therefore consider particle shape and orientation while sinking to be of relativelysmall influence since shape only becomes an important factor when width deviates considerably from length[McNown and Malaika, 1950].

We propagated assumed imprecision of 5% in Umeasured, 2% in ESD, 0.1% in ρseawater, and μseawater in equa-tion (2) to assess the statistical error in ρaggregate. This analysis suggests an error of around 0.0012 g cm�3 inρaggregate, which is small compared to absolute ρaggregate values. However, in the context of sinking velocity, itis necessary to consider changes in excess density (i.e., ρaggregate� ρseawater). For the latter, we estimated astatistical error of ~0.0016 g cm�3 based on the abovementioned individual imprecisions in ρaggregate andρseawater. This resulted in a relative error (i.e., 0.0016 g cm�3 divided by absolute excess density) between 6and 40%, which increased with decreasing ρaggregate.

Porosity is the fraction of an aggregate not occupied by solid matter [Alldredge and Gotschalk, 1988].Assuming that the investigated aggregate had no porosity, then its theoretical density (ρtheo) would be

ρtheo ¼ 1:06� fracPOC þ 2:1� fracBSi þ 2:7� fracPIC (3)

where fracPOC, fracBSi, and fracPIC are the relative contribution of each of the three materials to the totalweight (equation (1)). An aggregate with a theoretical density between 1.06 (pure POC) and 2.7 g cm�3 (pureCaCO3) will then have to be “diluted” with seawater (ρseawater 1.025 g cm

�3) to such a degree that ρtheo plusρseawater equals ρaggregate or

ρaggregate ¼ ρtheo þ u�ρseawateru þ 1

(4)

where u is the dilution factor. Solving equation (4) for u yields

u ¼ ρtheo � ρaggregateρaggregate � ρseawater

(5)

Porosity estimates (Pestimated) were subsequently calculated as

Pestimated ¼ 1� 1uþ 1

(6)

Potential imprecisions in density variables will also be reflected in the precision of Pestimated. We estimated astatistical imprecision of ~40 g cm�3 in ρtheo based on assumed imprecisions of 5% in the theoretical densityof POC [Bach et al., 2012a], as well as 2% in fracPOC, fracBSi, and fracPIC, respectively. This, together with theabovementioned error in ρaggregate and ρseawater lead to a statistical imprecisions between 2 and 30 in thedilution factor u (Δu). Lower Δu coincided with lower absolute values in u, which explains why the absoluteimprecision in Pestimated is decreasing with increasing Δu. For example, we estimated that Pestimated rangesbetween 0.985 and 0.992 when u is 100 and Δu is 30, while it ranges between 0.909 and 0.933 when u is12 and Δu is 2 (equation (6)). This analysis indicates that the imprecision in Pestimated is generally smaller thanthe changes observed in the course of the study, although it can be in the same range in some occasions(as exemplified in the second case described in the previous sentence). Aggregate fragmentation and



reassembly through sample treatment before measurements (see section 2.3) was another potential sourceof error in ρaggregate and hence Pestimated. This aspect will be addressed in section 4.2.2.

A critical assumption for the accuracy of the porosity calculation is the equality of chemical composition in allaggregate size classes. Deviations could for instance arise from large numbers of mesozooplankton in thetraps, which may bind an increased fraction of the POC in a particular size class. Mesozooplankton, however,was far less visible than phytodetritus, which strongly dominated the sediment trap material until the end ofthis investigation (appearance of sediment trap material shown in Figure S1 in the supporting information).Although we cannot entirely rule out the possibility of systematic changes in composition with size, there arethree lines of evidence indicating that this was rather not the case: (1) large aggregates are composed ofsmaller aggregate building blocks [Kiørboe, 2001; Burd and Jackson, 2009] so that variability in chemicalcomposition should average out unless the aggregate is composed of a single component [Alldredge,1998]; (2) sinking velocity and ρaggregate showed very similar trends in all four size classes, suggesting thatparticle properties do not systematically change with size (section 3.2); and (3) while filming the sinkingparticles with the microscope camera (FlowCam), we noted that the aggregate matrix appeared fairly homo-genous among the size classes (Figure S1).

2.6. Assessment of Transfer Efficiency

Transfer efficiency is defined as the ratio of sequestration flux to export flux, where the former is the amountof sinkingmatter reaching the bottom of themesopelagic zone (i.e., 1000m) and the latter the amount reach-ing the bottom of the euphotic zone (see Sanders et al. [2014] for definitions). Sinking velocities of surfaceocean aggregates as determined in this study cannot be used directly to derive transfer efficiencies. Wetherefore formulated a simple one-dimensional carbon flux model in order to test whether the magnitudeof change in measured sinking velocities could affect transfer efficiency.

Key assumptions for the one-dimensional model are (1) the depth of the euphotic zone is 100m and particlespassing this depth enter the export pathway, (2) lateral POC inputs or losses below the euphotic zone are bal-ancing each other, and (3) particle remineralization (kremin (day

�1)) during sinking is parameterized as a func-tion of temperature (T in °C) according to Schmittner et al. [2008]:

kremin ¼ 0:048 � 1:066T zð Þ (7)

where z is the depth in meters. This equation accounts for a doubling of enzymatic activity for a 10° increaseof temperature (Q10 kinetics). For T(z) we used the average profile at 61.5°N between 16 and 17°W from July2013 downloaded from the world ocean atlas data set (https://www.nodc.noaa.gov/OC5/indprod.html),which is an open ocean region but approximately the same latitude as our study site. (4) Model sinking velo-city (md�1) is parameterized as

Umodel ¼ 0:04z þ 25 (8)

The stepwise increase in Umodel is explained with a loss of the organic carbon content in aggregates [Berelson,2002]. Note that this parameterization was adopted from the University of Victoria model [Schmittner et al.,2008], but the sinking velocity at the surface was changed from 7md�1 [as in Schmittner et al., 2008] to25md�1 in order to receive a transfer efficiency of ~14% between 100 and 1000m depth in the controlrun, which corresponds to a Martin b of about 0.85 [Martin et al., 1987]. This value is around the average fluxattenuation among different ocean basins [Berelson, 2001]. (5) The POC fraction which is remineralized onevery depth interval depends on the balance between remineralization rates and sinking velocities. Withthese assumptions we can formulate the one-dimensional carbon flux model as

POC zð Þ ¼ POC z � 1ð Þ � POC z � 1ð Þ � kremin

Umodel� z � z � 1ð Þ½ �

� �(9)

where z� 1 is the depth interval above the calculated depth.

The sinking velocity term (Umodel; equation (8)) was then multiplied with a stepwise increasing factor in orderto simulate an acceleration of aggregate sinking speed over the entire water column. The correspondingincrease of transfer efficiency of POM (in percent) between the bottom of the euphotic zone (here 100m)and 1000m was calculated for each multiplication step as

Transfer efficiency ¼ POM1000

POM100(10)



https://www.nodc.noaa.gov/OC5/indprod.html

2.7. Data Analysis

The change of organic matter ballast-ing, aggregate sinking velocities, size,ρaggregate, and Pestimated over time wereanalyzed by means of time series analy-sis. Therefore, we averaged the dailyresponse in the respective parametersover all eight mesocosms. Calculation ofthe daily average smoothed the dataand facilitated the detection of trends.

Based on the trends observed in theaveraged aggregate parameters we splitthe whole data set into three distinctphases: phase I = Julian days 128–142,phase II = Julian days 143–152, and phaseIII = Julian days 153–157. We subse-quently tested for each time series para-meter whether there is a significanttrend (decrease or increase with time) inthe respective phase by means of linearregression analysis using R (R Project forStatistical Computing). Results of theseanalyses are summarized in Table S1 inthe supporting information. Please notethat we stopped our analysis on Julianday 157, although the study lasted untilday 163 because sinking velocity, thecore parameter in the present study,was only measured until day 157.

Temporal developments in zooplanktonabundance have not been analyzed withthe time series approach because oflow sampling frequency. The temporaldevelopment in phytoplankton abun-dance, chlorophyll a, and BSi was soobvious that a time series analysis wasnot necessary.

3. Results3.1. Plankton Community Structure

The mesocosms were closed during a postbloom period with rather low chlorophyll a and nutrient concen-trations but relatively high numbers of mesozooplankton. After closing the mesocosms the water columnwas gently mixed by air bubbling in order to eliminate the salinity stratification. This procedure lifted nutri-ents from deeper mesocosm water layers to the brighter top, thereby stimulating photoautotroph growthin phase I (Figure 2A). The increase in chlorophyll a was paralleled by decreasing nutrient concentrations(Figures 2B–2D) and primarily due to growth of picoeukaryotes (0.2–2μm), two groups of nanophytoplank-ton (NANO I (2–6μm) and NANO II (6–20μm)), and cryptophytes (2–20μm; Figure 3). The two NANO groupswere most likely the diatoms species Arcocellulus sp. and Thalassiosira sp. (R. Bermudez-Monsalve personalcommunications), which explains the increase in BSi (Figure 2E). Abundances of the coccolithophoreEmiliania huxleyi (5–10μm) and Synechococcus (0.6–1.6μm) remained low at the beginning (Figure 3).

Copepod abundance was highest at the beginning of the experiment and ranged from ~12.9 (M1) to 15.3(M7) individuals L�1. Roughly half of the animals were nauplius larvae (between 44 and 58%), but they

Figure 2. Temporal development of (A) chlorophyll a, (B) BSi, (C) NO3�,

(D) PO43�, and (E) Si(OH)4 concentrations. Julian day 127 was the 7th of

May. The vertical grey lines separate the three phases. NO3� and PO4

3�

were added on day 142.



developed to copepodites and adults during the first phytoplankton bloom (Figures 4A and 4C). Most of thecopepods (generally more than 95%) belonged to the genera Oithona, Temora, Pseudocalanus, and Calanusthroughout the entire experiment [Hildebrandt, 2014]. The loss of nauplii from the water column throughsedimentation was low but higher at the beginning than at any other time of the experiment (Figure 4B).This contrasted the loss rates of copepodites and adults, which were generally on a higher level but lowestat the beginning (Figure 4D). Metazoan calcifiers such as pteropods, other gastropod larvae, and bivalvelarvae (called mollusks in the following) were initially relatively abundant but quite constantly decreasedduring the entire experiment. Mollusk loss through sedimentation seemed to be higher at the beginningand toward the end of the experiment, although loss rates were difficult to quantify as there were largefluctuations between samplings (Figure 4F). Appendicularians, represented by the species Oikopleura dioica,were initially present in low numbers, and we observed an elevated loss through sedimentation at thebeginning of the study (Figures 4G and 4H).

The initial phytoplankton bloom started to decline between Julian days 131 and 132 (Figure 2A) mainly dueto a decreasing abundance of the PICO and NANO II groups (Figures 2C and 2D). The decline of NANO II wasparalleled by decreasing BSi concentrations (Figure 2B), decreasing copepod abundances in the watercolumn (Figures 4A and 4C), and increasing copepod loss rates through sedimentation (Figure 4D).

The addition of nutrients at the beginning of phase II (Figures 2B–2D) stimulated a second phytoplanktonbloom (Figure 2A), which was primarily mediated by NANO I, NANO II, and cryptophytes, except for M9

Figure 3. Temporal development of major phytoplankton functional types or genera. The vertical grey lines separate thethree phases. NO3

� and PO43� were added on day 142.



and to some extent also M7, where picoeukaryotes and Synechococcus were increasing instead (Figures 3B–3Dand 3F). As in the first bloom, autotrophic growth was paralleled by buildup of biogenic silica (Figure 2B).Copepod abundances during the second bloom were comparable to those before the nutrient fertilization(Figures 4A and 4C). Their loss rates through sedimentation, however, decreased by about 50% and stayed ata low level until Julian day 155 (Figure 4D). Appendicularians were still low in abundance but started to increasein phase II until the end of the experiment (Figure 4G).

The second bloom started to decline around Julian day 148, but the decline rate varied amongmesocosms. Itwas most pronounced in M1 and least pronounced in M9. Copepod and appendicularian abundance variedconsiderably among the mesocosms during the bloom decline. Their loss rates through sedimentation weregenerally low but increased when chlorophyll a concentrations reached baseline levels around Julianday 155.

Figure 4. Mesozooplankton development. (A, C, E, G) Measured water column abundances of the major mesozooplanktontypes. (B, D, F, H) Loss rates of these types through sedimentation. Note that the last measurement of zooplanktonabundance in the water column was on day 161, which was after sinking velocity measurements ended. The vertical greylines separate the three phases. NO3




Synechococcus and E. huxleyi strongly increased in abundances in some of the mesocosms toward the end ofthe experiment in phase III. Synechococcus was primarily blooming in M9 and M7 (up to 1.7 and0.5 × 106 cellsmL�1, respectively; Figure 3A), while E. huxleyi thrived most noticeably in M4, M6, and M1(up to 2800, 1900, and 1400 cellsmL�1, respectively; Figure 3E). The divergence of the plankton compositionbetween mesocosms seen in Figures 3 and 4 is the result of different CO2 concentrations. This will bediscussed in detail in publications currently in preparation.

3.2. Aggregates Properties and Sinking Velocity

Average sizes of aggregates within the four size classes were 101 (80–130μm), 145 (120–180μm), 205(170–260μm), and 289μm (240–400μm). Size was comparatively stable and showed little change over time(Figures 5A–5D), although it must be kept in mind that the particle size distribution in the settling columnpotentially deviated from the size distribution inside the mesocosms due to sample preparation[Boxhammer et al., 2016] (see also section 2.3).

The ρaggregate of the four size classes averaged over thewhole experiment decreasedwith aggregate size from1.044 (80–130μm), over 1.038 (120–180μm), 1.034 (170–260μm), to 1.031 g cm�3 (240–400μm). Theρaggregate decrease with size (Figure S2A) was due to the modular or fractal organization of aggregates[Alldredge and Gotschalk, 1988; Burd and Jackson, 2009]. Development of ρaggregate over time was comparableamong all size classes (Figures 5E–5H). Phase I was characterized by consistently high ρaggregate in the largertwo size classes and initially high but significantly decreasing ρaggregate in the two smaller ones(Figures 5E–5HandTable S1). Duringphase II ρaggregate suddenlydropped (larger two size classes) or continueddropping for 3 days (smaller two size classes) before reachingabaseline value, atwhich they remaineduntil theend of the second phase. The last phase was characterized by a treatment-specific development in ρaggregatewith those mesocosms with higher abundances of the coccolithophore E. huxleyi having higher values.

In the different size classes, Pestimated averaged over the whole experiment was 0.946 (80–130μm), 0.963(120–180μm), 0.976 (170–260μm), and 0.984 (240–400μm), which is in very good agreement with the results

Figure 5. Temporal development of aggregate properties and ballasting. (A–D) ESD, (E–H) ρaggregate, and (I–L) Pestimated of the aggregates within the four sizeclasses. (M–P) Contribution of POC, BSi, and PIC to the total weight (excluding seawater) of the material collected in the sediment traps. Y axis labels are shownon top of each plot. The black line is the mean calculated as the daily average of all mesocosms. The vertical grey lines separate the three phases. NO3

� and PO43�

were added on day 142.



from earlier studies [Alldredge andGotschalk, 1988; Lam and Bishop, 2007;Engel et al., 2009]. The increase ofPestimated with aggregate size is acommonly observed feature and dueto the fractal nature of aggregates[Alldredge and Gotschalk, 1988; Burdand Jackson, 2009]. Development ofPestimated over time was similar amongthe four size classes. It decreased signif-icantly throughout phase I (Table S1)but developed some treatment-specific differences toward the end ofthis phase (Figures 5I–5L). Differenceswere most pronounced in M9, whichgenerated aggregates with lowerPestimated. The treatment-specific differ-ences in Pestimated were still presentat the onset of phase II. However,Pestimated was starting to increase againin all mesocosms and reached a secondpeak shortly after the second peak inchlorophyll a near the end of phaseII (compare Figures 2A and 5I–5L).Pestimated decreased during the last2 days of phase II, but the decline wastreatment specific. It was again mostpronounced in M9 (Figures 5I–5L).

Aggregate sinking velocities averagedover the whole experiment were 6.1(80–130μm), 8.6 (120–180μm), 11.4(170–260μm), and 15.2 (240–400μm)md�1, respectively, and increased withaggregate size (Figure S2B). Changesin sinking velocity over time are similaramong the four size classes (Figure 6)reflected in the linear correlationbetween sinking velocities of differentsize classes (Figure S2C). Its temporaldevelopment mirrored changes inρaggregate (see above and compareFigures 6 and 5E–5H). Thus, all relevantchanges in aggregate sinking velocitymeasured in the settling columns werecaused by changes in their densityproperties.

3.3. Ballasting of Sediment Trap Material

At the beginning of the experiment roughly 65% of the total sediment material weight were made up byorganic matter, while silica and CaCO3 contributed 25 and 10%, respectively (Figures 5M–5P). The fractionof silica and CaCO3 decreased steadily until day 140 to ~12 and 7%, respectively. The continuous reductionof silica ballasting during this period was paralleled by the decline of diatoms and BSi in the water column(Figures 2B and 5N). The reduction of CaCO3 contribution in phase I was not reflected in a decrease of E.

Figure 6. (A–D) Measured sinking velocities of the four size classes (withFigure 6D being the smallest). The black line is the mean sinking velocitycalculated as the daily average of all mesocosms. The vertical grey linesseparate the three phases. NO3




huxleyi cell abundance (Figures 3E and 5O) but may instead either be attributable to (1) decreasing coccolithproduction rate at stable E. huxleyi population size, (2) the dwindling presence of calcifying mollusk larvae(Figure 4E), or (3) to a declining abundance of bacteria-derived CaCO3 precipitates [Heldal et al., 2012] (seealso section 4.1.1).

Silica contribution to total density started to increase again on day 146, which was 4 days after the nutrientaddition (Figure 5N). The increased share of silica material in the sediment trap of 15–25% shortly after thepeak of the second bloom (Julian day 150) coincides with a decline of water column BSi as well as NANO Iand NANO II abundances (Figures 2B, 3B, and 3D). Silica contribution decreased linearly after day 150 reflect-ing the decrease of BSi in the water column. Its contribution was 5–13% at the end of the experiment.

CaCO3 contribution to total density generally scaled positively with E. huxleyi abundance for the time afterthe nutrient addition until the end of the experiment. Highest E. huxleyi abundance in M4 translated to28% CaCO3 contribution to weight fraction of sediment trap material at the end of the experiment. Incontrast, CaCO3 was not contributing at all to the material weight in M9 during this phase (Figure 5O), whereE. huxleyi appeared in very low abundances (Figure 3E).

3.4. Relevance of Changes in Sinking Velocity for Particulate Matter Transfer Efficiency

Aggregate sinking velocities changed considerably in the course of the experiment in response to the plank-ton community succession (Figure 6). We formulated the one-dimensional carbon fluxmodel (equation (9)) inorder to assess whether changes within this scale could be relevant for particulate matter sequestration.Results of this sensitivity analysis are shown in Figures S3A and S3B. The pattern of particle remineralizationresembled a Martin curve, irrespective of the modeled sinking velocities. Flux attenuation was negativelycorrelated with sinking velocities (Figure S3A). Transfer efficiency between 100 and 1000m depth increasedby about 0.25% per 1% increase in modeled sinking velocity (Figure S3B).

4. Discussion

Linking plankton community structure with measurement of export-relevant parameters by means of in situmesocosm studies has recently been identified as a research priority [Sanders et al., 2014]. We enclosed a nat-ural plankton community in a Norwegian fjord and were able to follow its development for more than amonth under close to natural conditions (see video by Boxhammer et al. [2015] to get an impression onthe plankton community). This experiment is the first in a series of several (Swedish coast, 2013; Coast offthe Canary islands, 2014; Norwegian coast, 2015), where sinking velocity measurements with high temporalresolution [Bach et al., 2012a] were linked with plankton community structure within the mesocosms. We willtherefore not only address the major hypothesis and outcomes of the study (section 4.1) but also discusslimitations and illustrate potential improvements of the mesocosm approach (section 4.2).

4.1. Influence of the Plankton Community Structure on Aggregate Properties and Sinking Velocities4.1.1. Was There a Transition from Ballast to Packaging-Dominated Control on Sinking Velocity?The first phytoplankton bloom was characterized by high BSi concentrations in the water column and alarge copepod population developing from nauplii into adults. Aggregates sank relatively fast during thisperiod, which appears connected to a large CaCO3 and BSi ballast loading (Figure 5P) rather than lowPestimated (Figures 5I–5L).

Potential CaCO3-forming organisms present in the mesocosms at this time comprised E. huxleyi (Figure 3E),metazoan calcifiers (Figures 4E and 4F), and bacteria which can contribute up to ~0.3μmol kg�1 of inorganicCaCO3 as particles (1–100μm) in Raunefjord [Heldal et al., 2012] through the release of polyamines [Yasumotoet al., 2014]. We estimated CaCO3 supply by E. huxleyi based on the assumptions that each cell contributed atotal of 2 pmol CaCO3 (1 pmol within the coccosphere plus 1 pmol as detached coccoliths [Balch et al., 1993]).The presence of 52–185 cellsmL�1 during the first 7 days (Figure 3E) would result in a water column

coccolithCaCO3 standing stock of 0.1–0.37μmol L�1, whereas the corresponding CaCO3 accumulation rate inthe sediment traps ranged from 0.012 to 0.052μmol L�1 d�1 during this period (CaCO3 sedimentation datanot shown). We estimated that 8–37% loss of the coccolithCaCO3 standing stock per day could fully accountfor the bulk CaCO3 recovered from the sediment traps during the first 7 days. Such a loss rate could have beeneasily compensated by new coccolith production by the E. huxleyi population [Bach et al., 2012b] inside the



mesocosms. Thus, CaCO3 ballast recovered from the sediment traps during the first 7 days could have poten-tially been supplied to a large extent by E. huxleyi. The relevance of other CaCO3 sources remains unclear, butsince, for example, metazoan shells were sparse and have a lower carrying capacity for organic matter, theymay not have been particularly efficient ballast materials [Schmidt et al., 2014].

Opal ballast particles were generated by diatoms (mainly Arcocellulus sp. and Thallassiosira sp.), which werethe only silicifiers present in noticeable quantity by this time. Both diatom frustules and E. huxleyi coccolithsare small (~3–15μm), which allows them to be more homogenously incorporated into an organic aggregatematrix and serve as efficient ballast particles [Schmidt et al., 2014]. Hence, we assume that relatively fast sink-ing during the first bloom was primarily due to opal ballasting by diatoms and presumably CaCO3 ballastingby E. huxleyi.

The contribution of BSi and CaCO3 to the total weight of sediment trap material was rapidly decreasing from~27% (day 137) to 16% (day 140) after the first bloom. Sinking velocities, however, were not noticeablyaffected by the sudden decline in ballasting (Figure 6). The absence of a clear reduction of sinking velocitiescan be explained by declining Pestimated (i.e., increasing aggregate compactness), which compensated for thereduction of BSi and PIC ballasting during this time (Figures 5I–5L). Interestingly, the estimated shift towardless porous aggregates happened when chlorophyll a reached baseline levels and NO3

� and PO43� were

close to (NO3�) or within (PO4

3�) the detection limits (Figures 2A–2C). Therefore, we speculate that thedecrease in Pestimated could have been caused by a transition in the food web from a diatom-dominated foodweb fueled by upwelled nutrients (new production) toward one, which was increasingly dependent onremineralized nutrients (regenerative production). Presumably, the shift toward regenerative productionmay have intensified recycling and led to a more thorough removal of fresh and fluffy components fromthe POC standing stock. This hypothesis is supported by observations of higher POC and PON contents indiatom-derived marine snow compared to aggregates containing more decomposed components[Alldredge, 1998]. The preferential removal of fluffy components would lead to an accumulation of more den-sely packed material in the sinking material [Lam et al., 2011]. Additionally, this repackaging may have beenfurther amplified by copepods feeding on sediment trap material since we counted increasing numbers ofliving copepods in the sediment trap samples at the point when chlorophyll a concentrations consolidatedat a temporal minimum after the first bloom (compare Figures 2A and 4D). Organic matter became sparsein the water column at this time so that more and more copepods may have detected the sediment trapas an additional food source [Lampitt et al., 1993].4.1.2. Sinking Velocity Control Through Aggregate Porosity? The Potential Role of PlanktonCommunity StructureThe addition of nitrate and phosphate to all mesocosms (Figures 2B and 2C) on day 142 initiated a secondphytoplankton bloom. Diatoms were a prominent component in this second bloom (Figure 2B) as therewas still dissolved silicate left in the system (~0.5μmol L�1 on day 142). The loss rate of copepods throughsedimentation declined during bloom buildup, suggesting that the animals were attracted by the fresh foodgrowing in the water column and relied to a lesser extent on debris collected in the sediment trap.

Sinking velocities responded to these water column processes with a profound decline and reached experi-ment minima briefly after nutrient addition (Figure 6). To our surprise, these lowest velocities were sustainedthroughout the entire diatom bloom between days 142 and 152 even though increasing amounts of BSiballast were collected in the sediment traps with a ballasting potential close to that observed during the firstbloom (Figure 5N). The absence of a positive BSi ballast effect on sinking velocities during the diatom bloomcould have three different reasons. First, BSi ballast was not integrated into the aggregate matrix. This is unli-kely, however, considering that much of it originates from small diatoms (section 3.1), which should have ahigh potential to attach to organic matter due to their relatively high scavenging potential [Burd andJackson, 2009; Schmidt et al., 2014]. Second, the dominant diatoms coagulated within aggregates activelydownregulated their excess density in order to remain at higher light intensities during nutrient-fertilizedgrowth as seen for large chain-forming species [Moore and Villareal, 1996]. This explanation is also rather unli-kely, since large chains were hardly present in our study and decelerated sinking under nutrient-repletedconditions is a species-dependent, not a universal feature of diatoms [Bienfang et al., 1982]. Third, perhapsmost likely seems to be that high Pestimated (Figures 5I–5L) decelerated sinking velocity and thus overcom-pensated the positive influence of mineral ballast during that time. This conclusion is also supported by



our daily inspections of the sediment trap samples, where we noted that the collected material looked like afluffy mush rather than a consortium of individual particles.

Comparing the succession of Pestimated during bloom one (phase I) with that of bloom two (phase II + phase III)reveals a remarkably similar pattern. Pestimated was rather high in both blooms as long as the system was in a“new production state,” where biomass buildup is fueled by upwelled or added inorganic nutrients. Pestimated

decreased, however, when switching back to a “regenerative state” under low inorganic nutrient concentra-tions in the aftermath of a bloom (compare Figures 2B and 2C and Figures 5I–5L). These observations are in linewith a growing body of literature that export of aggregates generated during the typical diatom spring bloom isrelatively inefficient due to their high porosity (i.e., low packaging) [Francois et al., 2002; Lam and Bishop, 2007;Lam et al., 2011; Puigcorbé et al., 2015].

To further explore mechanisms controlling packaging we not only looked at the temporal development ofPestimated but also had a closer look on the differences among mesocosms. Differences were generally smalland the overall trend was similar, but we noticed that Pestimated tended to be lower in those mesocosms,which harbored higher numbers of picophytoplankton (e.g., M5, M7, and M9; Figures 3A and 3C). In orderto test this observation, we first calculated the biovolume of all phytoplankton groups (Figure 9), then calcu-lated the ratio of groups smaller 2μm (PICO) to groups larger 2μm in diameter (NANO), and thereafter cor-related the average PICO/NANO ratio with the average Pestimated (Figure 7). We found a negative correlationbetween the PICO/NANO ratio and Pestimated in all four aggregate size classes (Figure 7), which suggests thatplankton communities with abundant picophytoplankton tend to generate more tightly packed aggregatesthan communities where a higher fraction of the biomass is accumulated in larger species. This observation is

Figure 7. Influence of plankton community structure on Pestimated. Estimated cell diameters used for biovolume calcula-tions (assuming spherical shape) were Synechococcus = 1 μm, picoeukaryotes = 1 μm, NANO I group = 3 μm, NANO IIgroup = 10 μm, E. huxleyi = 6 μm, and cryptophytes = 7 μm. Biovolume was subsequently calculated by multiplication offlow cytometry counts (Figure 3) with species-specific biovolumes. The sum of biovolume of species smaller than 2 μm(Synechococcous and picoeukaryotes) divided by the sum of biovolume provided by species larger than 2 μm (NANO I,NANO II, E. huxleyi, and cryptophytes) is the PICO:NANO ratio. Shown here is the PICO:NANO ratio and Pestimated averagedover the whole experiment with one line for each of the four aggregate size classes.



mechanistically reasonable because small aggregate components allow relatively little interstitial spacebetween them [Burd and Jackson, 2009]. Furthermore, PICO-dominated communities usually prevail inregenerative systems, where particles potentially undergo a relatively large degree of biotic reprocessingand therefore a continuous compression [Lam et al., 2011]. With respect to particle matter export, lowPestimated aggregates formed in a high PICO/NANO regime would promote transfer efficiencies as they prob-ably sink faster than fluffy aggregates. Thus, the negative correlation between PICO/NANO and Pestimated

(Figure 7) observed in our study may help to explain the numerous observations that low productivityregimes tend to generate particles with high transfer efficiencies [Francois et al., 2002; Lam and Bishop,2007; Lomas et al., 2010; Lam et al., 2011; Henson et al., 2012a, 2012b; Maiti et al., 2013; Cavan et al., 2015;Puigcorbé et al., 2015].4.1.3. The Potential for Sinking Velocity Acceleration by E. Huxleyi-Derived CaCO3 BallastPhase III was characterized by decreasing chlorophyll a and BSi concentrations as well as treatment-specificdevelopments within major phytoplankton and zooplankton types (Figures 3 and 4). The differential devel-opments amongmesocosms were reflected in particle sinking velocities, where those mesocosms which har-bored significant E. huxleyi blooms generated aggregates with some of the highest sinking velocitiesmeasured during the entire study, while aggregate sinking velocities remained at a low level in mesocosmswhere E. huxleyi growth was muted (Figures 3 and 6).

The acceleration of sinking velocities in phase III was most likely due to CaCO3 ballast provided by E. huxleyi(Figure 8) and rather not caused by tight packaging because Pestimated was higher in those treatments withhigh CaCO3 availability (compare Figures 5I–5L with 5O). We used the positive correlations between particlesinking velocity and E. huxleyi abundance (Figure 8A) or sediment material PIC:POC ratio during phase III(Figure 8D) to estimate the influence of E. huxleyi-dependent acceleration of particle sinking velocity on trans-fer efficiency with the one-dimensional particle flux model (equation (9)) derived in section 2.6. Therefore,slopes and the y intercepts of the four sinking velocities versus PIC:POC or cell abundance correlations (seeFigures 8A and 8D) were plotted against average ESD of the respective size class. Slopes and y interceptsof the four correlations increased with average aggregate size (Figures 8B, 8C, 8E, and 8F), which is expectedaccording to Stokes’ law. With the slopes versus ESD (Figures 8B and 8E) and y intercepts versus ESD

Figure 8. Influence of the E. huxleyi blooms on sinking velocities during phase III. (A) Sinking velocities of the four size classes (black triangles: 80–130 μm; blue dots:120–180 μm; green squares: 170–260 μm; red pyramids: 240–400 μm) as a function of water column E. huxleyi abundance. Each data point represents the phase IIIaverage of one mesocosm with error bars being standard deviations. Note that we calculated phase III averages (instead of using daily measurements during phaseIII) as we are unable to tell with what temporal delay E. huxleyi cells arrive in the sediment traps. (B) Correlations between slopes and average ESD or (C) y interceptsand average ESD of the regressions calculated in Figure 8A. (D–F) The same results as in Figures 8A–8C, but in this case, sinking velocities are a function of PIC:POCratio measured in sediment trap material. Note that data from individual days during phase III could be used in Figure 8D in contrast to Figure 8A because mea-surements of PIC:POC and sinking velocities were done with material collected on the same day.



(Figures 8C and 8F) correlations, we were able to parameterize the dependency of sinking velocities on eithersediment material PIC:POC ratio or water column E. huxleyi abundance (Nehux) as

SinkingvelocityPIC:POC ¼ slopePIC:POC�PIC : POCþ interceptPIC:POC (11)

and

Sinkingvelocityehux ¼ slopeehux�Nehux þ interceptehux (12)

where slopePIC:POC and slopeehux as well as interceptPIC:POC and interceptehux were taken from regressionsshown in Figures 8B, 8C, 8E, and 8F. The E. huxleyi- or PIC:POC-dependent acceleration of sinking velocitiescan subsequently be included in the export model (equation (9)) by multiplying the sinking velocitymodel

term with the ehuxfactor or the PIC:POCfactor, defined as

ehuxfactor ¼ slopeehux�Nehux þ interceptehuxinterceptehux

(13)

and

PIC : POCfactor ¼ slopePIC:POC�PIC : POCþ interceptPIC:POCinterceptPIC:POC

(14)

Hence, Umodel in equation (9) increases about a certain percentage throughout the entire water column,similar as has been simulated in the sensitivity analysis conducted in section 3.4. This time, however, sinkingvelocity is not increased by an arbitrary factor but scales with E. huxleyi cell concentrations in the euphoticzone or the PIC:POC ratio of the sinking aggregates.

Sinking velocity acceleration was already profound under relatively low E. huxleyi abundances or PIC:POCratios (Figures 8A and 8D), which also considerably increased transfer efficiencies (Figure 9). For example,we estimated an acceleration of almost 40% for an increase in E. huxleyi abundance from 0 to1500 cellsmL�1 or a PIC:POC increase of ~0.1 (Figures 8A and 8D). Such a ballast-mediated increase in sinkingvelocity could elevate transfer efficiencies from 14 to 24% (Figure 9). The key question now is: Are E. huxleyicell densities of 1500 cellsmL�1 and the corresponding increase in PIC:POC exceptional or common in the

Figure 9. Influence of the E. huxleyi blooms on transfer efficiencies during phase III calculated with the one-dimensionalcarbon flux model (section 4.1.3). (A) Flux attenuation as a function of different E. huxleyi-dependent sinking velocityparameterizations (section 4.1.3; Figure 8a). The upper and lower horizontal grey lines indicate the bottom of the euphoticzone and sequestration depth, respectively. (B) Increase of transfer efficiency with E. huxleyi abundance in the euphoticzone. (C and D) The same results as in Figures 9A and 9B, but in this case, sinking velocities are a function of PIC:POC ratiomeasured in sediment trap material (section 4.1.3; Figure 8D).



oceans? Abundances of up to 1000 [Poulton et al., 2013], 1500 [Balch et al., 1991], and even 20,000coccospheresmL�1 [Holligan et al., 1993] were counted during bloom seasons on the Patagonian Shelf,the Gulf of Maine, and south of Iceland (63°N, 20°W), respectively. Kopelevich et al. [2014] calculated that E.huxleyi often reaches abundances of 3000 cellsmL�1 or even more during the typical coccolithophore sum-mer bloom in the Black Sea. An estimated 3mmol PICm�3 contributed by 1500 cellsmL�1 (assuming2 pmol PIC cell�1 [Balch et al., 1993]) also compares reasonably well with data from the satellite studies byMoore et al. [2012] and Hopkins et al. [2015], who found coccolithophore PIC of 1–10mmolm�3 to be com-mon on shelves and many regions of the open ocean during bloom season. Based on these assessmentson E. huxleyi abundance we conclude that E. huxleyi-associated particle sinking velocity accelerations of~40% or even more are perhaps not unusual. Thus, the regional and/or seasonal occurrence of E. huxleyiblooms may mark events of highly efficient sequestration pulses.

4.2. Sinking Velocity Measurements of Mesocosm Sediment Trap Material: Uncertainties, Limitations,and Future Directions4.2.1. Influence of Lithogenic Ballast MaterialsSome of the highest sinking velocities of the entire experiment were measured during the first days(Figure 6). As lithogenic material has not been quantified in this study, it cannot be ruled out that the initiallyfast sinking rates are due to lithogenic ballast minerals enclosed while lowering the mesocosm bags at thebeginning of the experiment. However, mesocosm bags were closed 4 days before the first sinking velocitymeasurement and the sediment traps were emptied on a daily basis until then. Artificial seeding of meso-cosms with Saharan dust particles (99% <1μm in size) showed that fastest sinking dust particles are trans-ferred to 9.5m within hours [Bressac et al., 2012]. Smaller particles remained longer in the water columnbut also declined considerably within 2 days [Bressac et al., 2012]. Thus, if lithogenic ballast particles were pre-sent in relevant quantities at the beginning of our experiment, then most of it should have sunk out of thewater column before the first sinking velocity measurement.

The Icelandic volcano Grímtsvötn erupted in the course of the experiment on 21 May. Part of the Grímtsvötnash plume passed the Raunefjord area during the night from 24 to 25 May (days 144–145) with ash concen-trations ranging from 200 to 4000μgm�3 air and ash particle sizes between ~1μm and 50μm [Tesche et al.,2012; Lieke et al., 2013]. Sinking velocities, however, remained low after the ash passage (Figure 5) for at leastanother 4 days, which is too long considering that particles within this size class should have arrived muchearlier in the sediment trap. Furthermore, the acceleration of sinking velocity 4 days after the ash passagewas inconsistent among mesocosms, which is in contrary to expectations in case of such a dust-seedingevent. Thus, it is concluded that dust from the Grímtsvötn eruption was unlikely to have a significant ballasteffect on mesocosm aggregates, either because ash concentrations were too low or because they were insuf-ficiently collected by the mesocosms.

Based on these two lines of evidence we conclude that lithogenic ballast materials did not play a crucial rolein the present study. Nevertheless, it is advisable to quantify their influence in mesocosm experiments onparticle export because they may play an important role in other settings, where dust influx is more likely.4.2.2. Sampling-Associated Limitations and Potential ImprovementsPrior to sinking velocity determination, aggregates had to be pumped from the sediment trap to the surface,carried to land in a bottle, and prepared for the measurements (see section 2.3). This procedure provokedaggregate fragmentation and reassembly, which potentially influenced aggregate porosity and densityand likely influenced aggregate size. Potential effects on porosity and density should be relatively constantover time and thus of minor consequence on the trends observed in this study. The likely modification ofthe aggregate size structure prevented us from investigating if changes in the plankton community compo-sition could influence aggregate sinking velocity through changes in aggregate size structure.

Our method to measure sinking velocities (as set up in this study) is restricted to particles from 80 to 400μmin ESD [Bach et al., 2012a]. Although particles of that size are responsible for a large fraction of organic matterexport to depth, we missed the important influence of particles outside that range [Clegg and Whitfield, 1990;Ebersbach and Trull, 2008; Durkin et al., 2015]. In our study it is likely, however, that sinking velocities of aggre-gates>400μmwould have shown a similar response to changing plankton community structure as the onesbelow this threshold as long as they are composed of smaller aggregate building blocks. The same principleshould apply for aggregates <80μm but only until they become so small that they reach the size of



individual components (e.g., single-diatom cells), where their specific chemical composition no longeraverages out in the aggregate matrix. The good correlation between sinking velocities of aggregates fromdifferent size classes (Figure S2C) supports the notion that temporal trends observed between 80 and400μm could to some extent also be found outside this size range.

The focus on rather small aggregates in this study (80–400μm) explains comparatively slow sinking rates,ranging from an average of 6md�1 in the smallest size class (80–120μm) to 15md�1 in the largest one(240–400μm). Extrapolation based on the size versus velocity correlation given in Figure S2B projects settlingrates of about 200md�1 for 4mm large aggregates. This aligns well with in situ measurements yielding1.3–287md�1 at a particle size of 0.2–20mm [Diercks and Asper, 1997; Pilskaln et al., 1998; Nowald et al., 2009]and in situ sensor estimates yielding 10–200md�1 at a size of 0.07–6mm [Mcdonnell and Buesseler, 2010;Jackson et al., 2015]. They tend to be somewhat lower, however, than sinking velocities calculated from sedimenttrap deployments with either the “benchmark”method (average of 70 to 450md�1 [Berelson, 2002; Fischer andKarakaş, 2009; Xue and Armstrong, 2009]) or specialized settling velocity sediment traps (ranging from 1 to2000md�1 [Armstrong et al., 2009; Alonso-González et al., 2010]). Potential explanations for discrepanciesbetween direct measurements and sediment trap-derived estimates could be poor trapping efficiency of slowsinking particles [Baker et al., 1988], leading to their underrepresentation in the sediment trap sample or remi-neralization of slow sinking particles before reaching the usually quite deeply deployed sediment traps.

Aggregate remineralization is the other factor (next to sinking velocity) controlling the attenuation of POM fluxwith depth (equation (9)). In the present study we neglected remineralization rate but we assume that changesin this parameter are opposite to the changes in sinking velocity. For example, tightly packed aggregates willsink faster than similar-sized fluffy aggregates but their remineralization will probably be slower because lessporous particles have a reduced surface to volume ratio. Likewise, ballast-mediated acceleration of sinking velo-cities can be accompanied by reduced remineralization rates in case the ballast encloses and physically protectsthe organic matter [Francois et al., 2002]. Thus, we speculate that the calculated trends in transfer efficiencieswould be amplified rather than buffered by changes in remineralization rates.4.2.3. Restriction to Surface Ocean Processes: The Need for “Meso-Pelagi-Cosms”The water column enclosed in the mesocosms was 25m deep. Accordingly, the time from particle formation inthe water column to collection in the sediment trap is relatively short and particles experienced little modifica-tions after production. Our analysis is therefore restricted to processes taking place in the epipelagic but leavespostproduction particle modifications within the mesopelagic realm unconsidered. Thus, sinking velocitiesreported here should be seen as “entry speed” of aggregates at the beginning of their descent into the deep.

This limitation could be solved by designing significantly longer mesocosm bags than the ones used here—ideally ones which reach hundreds of meters deep into the mesopelagic. These Meso-pelagi-cosms wouldfully exclude advection and therefore enable us to understand and precisely quantify organic matter fluxesinto the deep, how they depend on the plankton community in the euphotic zone, and how they are attenu-ated below the euphotic zone. Although construction and successful deployment of such Meso-pelagi-cosmsare technically challenging, it may be worth aiming for such an approach in the future.

5. Conclusions

In this study we have investigated the influence of community structure on aggregate sinking velocity. We con-clude that plankton communities influence sinking velocities primarily by providing ballast minerals andmodify-ing aggregate porosity, although it must be kept in mind that the porosity estimate in this study (Pestimated) is ahighly derived parameter and therefore associated with uncertainties (sections 2.5 and 4.2.2). Our observationspoint toward the following: (1) Ballast availability was relatively high when phytoplankton blooms were fueledby inorganic nutrients and dominated by diatoms. Yet sinking velocities of aggregates generated during thesebloom were not necessarily accelerated by ballast. This could be related to the relatively high porosity of aggre-gates produced when the plankton community is in a new production state. (2) Pestimated tended to decrease inpostbloom periods, which we hypothesize to be caused by a shift from new to regenerated production. (3)Pestimated averaged over the entire study period was lower in those mesocosms, where picophytoplanktoncontributed a larger fraction to the autotrophic biovolume. Therefore, PICO-dominated communities may beindicative for foodweb structures promoting a high degree of particle repackagingwith potential for acceleratedsinking velocities. Our Pestimated-based interpretations (2, 3) align very well with the growing body of literature



pronouncing the important role of particle repackaging in particulate matter export [e.g., Lam et al., 2011;Hensonet al., 2012b; Puigcorbé et al., 2015]. (4) Some of the highest sinking velocities were measured during a moderateE. huxleyi bloom toward the end of the experiment. This is in line with the common notion that CaCO3 is a parti-cularly effective ballast mineral and emphasizes the importance of E. huxleyi to support regional organic mattersequestration pulses.

Author Contributions

U.R. conceived, initiated, and organized the SOPRAN 2011 mesocosm study. L.T.B., T.B., and K.G.S. developedthe subproject to study export of sinkingmaterial by means of sinking velocity measurements. L.T.B., T.B., A.L.,N.H., and K.G.S. analyzed the data. L.T.B. interpreted the data and wrote the paper with comments fromall coauthors.

ReferencesAlldredge, A. (1998), The carbon, nitrogen andmass content of marine snow as a function of aggregate size, Deep Sea Res., Part I, 45, 529–541,

doi:10.1016/S0967-0637(97)00048-4.Alldredge, A. L., and C. Gotschalk (1988), In situ settling behavior of marine snow, Limnol. Oceanogr., 33(3), 339–351, doi:10.4319/lo.1988.33.3.0339.Alonso-González, I. J., J. Arístegui, C. Lee, A. Sanchez-Vidal, A. Calafat, J. Fabrés, P. Sangrá, P. Masqué, A. Hernández-Guerra, and

V. Benítez-Barrios (2010), Role of slowly settling particles in the ocean carbon cycle, Geophys. Res. Lett., 37, L13608, doi:10.1029/2010GL043827.

Armstrong, R. A., C. Lee, J. I. Hedges, S. Honjo, and S. G. Wakeham (2002), A new, mechanistic model for organic carbon fluxes in the oceanbased on the quantitative association of POC with ballast minerals, Deep Sea Res., Part II, 49(1-3), 219–236.

Armstrong, R. A., M. L. Peterson, C. Lee, and S. G. Wakeham (2009), Settling velocity spectra and the ballast ratio hypothesis, Deep Sea Res.,Part II, 56(18), 1470–1478, doi:10.1016/j.dsr2.2008.11.032.

Bach, L. T., U. Riebesell, S. Sett, S. Febiri, P. Rzepka, and K. G. Schulz (2012a), An approach for particle sinking velocity measurements in the3–400 μm size range and considerations on the effect of temperature on sinking rates, Mar. Biol., 159(8), 1853–1864, doi:10.1007/s00227-012-1945-2.

Bach, L. T., C. Bauke, K. J. S. Meier, U. Riebesell, and K. G. Schulz (2012b), Influence of changing carbonate chemistry on morphology andweight of coccoliths formed by Emiliania huxleyi, Biogeosciences, 9, 3449–3463, doi:10.5194/bg-9-3449-2012.

Baker, E. T., H. B. Milburn, and D. A. Tennant (1988), Field assessment of sediment trap efficiency under varying flow conditions, J. Mar. Res.,46, 573–592, doi:10.1357/002224088785113522.

Balch, W. M., P. M. Holligan, S. G. Ackleson, and K. J. Voss (1991), Biological and optical properties of mesoscale coccolithophore blooms in theGulf of Maine, Limnol. Oceanogr., 36(4), 629–643, doi:10.4319/lo.1991.36.4.0629.

Balch, W. M., K. Kilpatrick, P. M. Holligan, and T. Cucci (1993), Coccolith production and detachment by Emiliania huxleyi (Prymnesiophyceae),J. Phycol., 29(5), 566–575.

Berelson, W. M. (2001), The flux of particulate organic carbon into the ocean interior: A comparison of four U.S. JGOFS regional studies,Oceanography, 14(4), 59–67.

Berelson, W. M. (2002), Particle settling rates increase with depth in the ocean, Deep Sea Res., Part II, 49, 237–251, doi:10.1016/S0967-0645(01)00102-3.

Bienfang, P. K., P. J. Harrison, and L. M. Quarmby (1982), Sinking rate response to depletion of nitrate, phosphate and silicate in four marinediatoms, Mar. Biol., 67, 295–302.

Boxhammer, T., L. T. Bach,M.Nicolai, andU. Riebesell (2015), Videoof aplankton community enclosed in a “KielOff-ShoreMesocosms for futureOcean Simulations” (KOSMOS) during the SOPRAN study in Raunefjord (Norway) 2011, doi:10.3289/KOSMOS_PLANKTON_NORWAY_2011.

Boxhammer, T., L. T. Bach, J. Czerny, and U. Riebesell (2016), Technical Note: Sampling and processing of mesocosm sediment trap materialfor quantitative biogeochemical analysis, Biogeosciences, 13, 2, doi:10.5194/bgd-12-18693-2015.

Boyd, P. W., and T. W. Trull (2007), Understanding the export of biogenic particles in oceanic waters: Is there consensus?, Prog. Oceanogr.,72(4), 276–312, doi:10.1016/j.pocean.2006.10.007.

Bressac, M., C. Guieu, D. Doxaran, F. Bourrin, G. Obolensky, and J. M. Grisoni (2012), A mesocosm experiment coupled with opticalmeasurements to assess the fate and sinking of atmospheric particles in clear oligotrophic waters, Geo-Mar. Lett., 32, 153–164,doi:10.1007/s00367-011-0269-4.

Burd, A. B., and G. A. Jackson (2009), Particle aggregation, Annu. Rev. Mar. Sci., 1, 65–90, doi:10.1146/annurev.marine.010908.163904.Cavan, E. L., F. A. C. Le Moigne, A. J. Poulton, G. A. Tarling, P. Ward, C. J. Daniels, G. M. Fragoso, and R. J. Sanders (2015), Attenuation of

particulate organic carbon flux in the Scotia Sea, Southern Ocean, is controlled by zooplankton fecal pellets, Geophys. Res. Lett., 42,821–830, doi:10.1002/2014GL062744.

Clegg, S. L., and M. Whitfield (1990), A generalized model for the scavenging of trace metals in the open ocean—I. Particle cycling, Deep-SeaRes., Part A, 37(5), 809–832.

Diercks, A.-R., and V. L. Asper (1997), In situ settling speeds of marine snow aggregates below the mixed layer: Black Sea and Gulf of Mexico,Deep Sea Res., Part I, 44(3), 385–398, doi:10.1016/S0967-0637(96)00104-5.

Durkin, C. A., M. L. Estapa, and K. O. Buesseler (2015), Observations of carbon export by small sinking particles in the upper mesopelagic,Mar.Chem., 175, 72–81, doi:10.1016/j.marchem.2015.02.011.

Ebersbach, F., and T. W. Trull (2008), Sinking particle properties from polyacrylamide gels during the KErguelen Ocean and Plateau comparedStudy (KEOPS): Zooplankton control of carbon export in an area of persistent natural iron inputs in the Southern Ocean, Limnol. Oceanogr.,53(1), 212–224, doi:10.4319/lo.2008.53.1.0212.

Engel, A., J. Szlosek, L. Abramson, Z. Liu, and C. Lee (2009), Investigating the effect of ballasting by CaCO3 in Emiliania huxleyi: I. Formation,settling velocities and physical properties of aggregates, Deep Sea Res., Part II, 56, 1396–1407, doi:10.1016/j.dsr2.2008.11.027.

Field, C. B., M. J. Behrenfeld, J. T. Randerson, and P. G. Falkowski (1998), Primary production of the biosphere: Integrating terrestrial andoceanic components, Science, 281(5374), 237–240, doi:10.1126/science.281.5374.237.



AcknowledgmentsWe thank all participants of the 2011SOPRAN mesocosm study, in particular,Signe Koch-Klavsen and Peter Fritzschefor measuring the chlorophyll a, BSi, andnutrient concentrations; Andrea Ludwigfor the support on logistics andorganization of the study; and ScarlettSett, Jan Büdenbender, Jan Czerny,Michael Sswat, Sonja Endres, LuisaGalgani, Hendrik Schultz, ChristianHowe, Hubert Kraus, Allanah Paul, AljosaZavisic, Allison Webb, Rafael Bermudez,Jorun Egge, Barbara Niehoff, DanielaFreese, and Michael Meyerhöfer fortheir technical support. We are gratefulto the crews of R/V Alkor (AL376), R/VHåkan Mosby (2011609), and R/VHeinke (HE360) for the mooring andtransportation of the mesocosms. Thispaper profited from the constructivecomments from two anonymousreviewers. The project was funded byGerman Ministry for Science andEducation (BMBF) in the framework ofthe SOPRAN and BIOACID projects withadditional support (salary Aud Larsen)from the projects MicroPolar (RCNproject 225956/E10) and MINOS(EU-ERC 250254). All data used here willbe made available on www.pangaea.deand can also be requested directly fromthe corresponding author ([email protected]).

http://dx.doi.org/10.1016/S0967-0637(97)00048-4

http://dx.doi.org/10.4319/lo.1988.33.3.0339

http://dx.doi.org/10.1029/2010GL043827

http://dx.doi.org/10.1029/2010GL043827

http://dx.doi.org/10.1016/j.dsr2.2008.11.032

http://dx.doi.org/10.1007/s00227-012-1945-2

http://dx.doi.org/10.1007/s00227-012-1945-2

http://dx.doi.org/10.5194/bg-9-3449-2012

http://dx.doi.org/10.1357/002224088785113522

http://dx.doi.org/10.4319/lo.1991.36.4.0629

http://dx.doi.org/10.1016/S0967-0645(01)00102-3

http://dx.doi.org/10.1016/S0967-0645(01)00102-3

http://dx.doi.org/10.3289/KOSMOS_PLANKTON_NORWAY_2011

http://dx.doi.org/10.5194/bgd-12-18693-2015

http://dx.doi.org/10.1016/j.pocean.2006.10.007

http://dx.doi.org/10.1007/s00367-011-0269-4

http://dx.doi.org/10.1146/annurev.marine.010908.163904

http://dx.doi.org/10.1002/2014GL062744

http://dx.doi.org/10.1016/S0967-0637(96)00104-5

http://dx.doi.org/10.1016/j.marchem.2015.02.011

http://dx.doi.org/10.4319/lo.2008.53.1.0212


http://dx.doi.org/10.1126/science.281.5374.237

http://www.pangaea.de



Fischer, G., and G. Karakaş (2009), Sinking rates and ballast composition of particles in the Atlantic Ocean: Implications for the organic carbonfluxes to the deep ocean, Biogeosciences, 6, 85–102, doi:10.5194/bg-6-85-2009.

Francois, R., S. Honjo, R. Krishfield, and S. Manganini (2002), Factors controlling the flux of organic carbon to the bathypelagic zone of theocean, Global Biogeochem. Cycles, 16(4), 1087, doi:10.1029/2001GB001722.

Giering, S. L. C., et al. (2014), Reconciliation of the carbon budget in the ocean’s twilight zone, Nature, 507(7493), 480–483, doi:10.1038/nature13123.

Hansell, D. A., and C. A. Carlson (2001), Marine dissolved organic matter and the carbon cycle, Oceanography, 14(4), 41–49.Hansen, H. P., and F. Koroleff (1999), Determination of nutrients, in Methods of Seawater Analysis, edited by K. Grasshoff, K. Kremling, and

M. Ehrhardt, pp. 159–226, Wiley-VCH, Weinheim.Hedges, J. I., and J. M. Oades (1997), Comparative organic geochemistries of soils and marine sediments, Org. Geochem., 27(7-8), 319–361.Heldal, M., S. Norland, E. S. Erichsen, T. F. Thingstad, and G. Bratbak (2012), An unaccounted fraction of marine biogenic CaCO3 particles, PLoS

One, 7(10), 1–6, doi:10.1371/journal.pone.0047887.Henson, S. A., R. Sanders, E. Madsen, P. J. Morris, F. Le Moigne, and G. D. Quartly (2011), A reduced estimate of the strength of the ocean’s

biological carbon pump, Geophys. Res. Lett., 38, 10–14, doi:10.1029/2011GL046735.Henson, S. A., R. Sanders, and E. Madsen (2012b), Global patterns in efficiency of particulate organic carbon export and transfer to the deep

ocean, Global Biogeochem. Cycles, 26, GB1028, doi:10.1029/2011GB004099.Henson, S., R. Lampitt, and D. Johns (2012a), Variability in phytoplankton community structure in response to the North Atlantic Oscillation

and implications for organic carbon flux, Limnol. Oceanogr., 57(6), 1591–1601, doi:10.4319/lo.2012.57.6.1591.Hildebrandt, N. (2014), The response of three dominant Arctic copepod species to elevated CO2 concentrations and water temperatures PhD

thesis, 1–173 pp., Alfred-Wegener-Institut, Helmholtz-Zentrum für Polar- und Meeresforschung, Bremen.Holligan, P. M., et al. (1993), A biogeochemical study of the coccolithophore, Emiliania huxleyi, in the North Atlantic, Global Biogeochem.

Cycles, 7(4), 879–900, doi:10.1029/93GB01731.Honda, M. C., and S. Watanabe (2010), Importance of biogenic opal as ballast of particulate organic carbon (POC) transport and existence of

mineral ballast-associated and residual POC in the Western Pacific Subarctic Gyre, Geophys. Res. Lett., 37, L02605, doi:10.1029/2009GL041521.

Honjo, S., S. J. Manganini, R. A. Krishfield, and R. Francois (2008), Particulate organic carbon fluxes to the ocean interior and factors controllingthe biological pump: A synthesis of global sediment trap programs since 1983, Prog. Oceanogr., 76(3), 217–285, doi:10.1016/j.pocean.2007.11.003.

Hopkins, J., S. A. Henson, S. C. Painter, T. Tyrrell, and A. J. Poulton (2015), Phenological characteristics of global coccolithophore blooms,Global Biogeochem. Cycles, 29, 239–253, doi:10.1002/2014GB004919.

Jackson, G. A., D. M. Checkley, and M. Dagg (2015), Settling of particles in the upper 100 m of the ocean detected with autonomous profilingfloats off California, Deep Sea Res., Part I, 99, 75–86, doi:10.1016/j.dsr.2015.02.001.

Kiørboe, T. (2001), Formation and fate of marine snow: Small-scale processes with large-scale implications, Sci. Mar., 65(2), 57–71.Klaas, C., and D. E. Archer (2002), Association of sinking organic matter with various types of mineral ballast in the deep sea: Implications for

the rain ratio, Global Biogeochem. Cycles, 16(4), 1116, doi:10.1029/2001GB001765.Kopelevich, O., V. Burenkov, S. Sheberstov, S. Vazyulya, M. Kravchishina, L. Pautova, V. Silkin, V. Artemiev, and A. Grigoriev (2014), Satellitemonitoring

of coccolithophore blooms in the Black Sea from ocean color data, Remote Sens. Environ., 146, 113–123, doi:10.1016/j.rse.2013.09.009.Lam, P. J., and J. K. B. Bishop (2007), High biomass, low export regimes in the Southern Ocean, Deep Sea Res., Part II, 54, 601–638, doi:10.1016/

j.dsr2.2007.01.013.Lam, P. J., S. C. Doney, and J. K. B. Bishop (2011), The dynamic ocean biological pump: Insights from a global compilation of particulate

organic carbon, CaCO3, and opal concentration profiles from the mesopelagic, Global Biogeochem. Cycles, 25, GB3009, doi:10.1029/2010GB003868.

Lampitt, R. S., K. F. Wishner, C. M. Turley, and M. V. Angel (1993), Marine snow studies in the Northeast Atlantic Ocean: Distribution,composition and role as a food source for migrating plankton, Mar. Biol., 116, 689–702, doi:10.1007/BF00355486.

Larsen, A., T. Castberg, R. A. Sandaa, C. P. D. Brussaard, J. Egge, M. Heldal, A. Paulino, R. Thyrhaug, E. J. Van Hannen, and G. Bratbak (2001),Population dynamics and diversity of phytoplankton, bacteria and viruses in a seawater enclosure, Mar. Ecol. Prog. Ser., 221, 47–57,doi:10.3354/meps221047.

Le Moigne, F. A. C., K. Pabortsava, C. L. J. Marcinko, P. Martin, and R. J. Sanders (2014), Where is mineral ballast important for surface export ofparticulate organic carbon in the ocean?, Geophys. Res. Lett., 41, 1–9, doi:10.1002/2014GL061678. Received

Lieke, K. I., T. B. Kristensen, U. S. Korsholm, J. H. Sørensen, K. Kandler, S. Weinbruch, D. Ceburnis, J. Ovadnevaite, C. D. O’Dowd, and M. Bilde(2013), Characterization of volcanic ash from the 2011 Grímsvötn eruption by means of single-particle analysis, Atmos. Environ., 79,411–420, doi:10.1016/j.atmosenv.2013.06.044.

Lomas, M. W., D. K. Steinberg, T. Dickey, C. A. Carlson, N. B. Nelson, R. H. Condon, and N. R. Bates (2010), Increased ocean carbonexport in the Sargasso Sea is countered by its enhanced mesopelagic attenuation, Biogeosciences, 7, 57–70, doi:10.5194/bgd-6-9547-2009.

Longhurst, A., S. Sathyendranath, T. Platt, and C. Caverhill (1995), An estimate of global primary production in the ocean from satelliteradiometer data, J. Plankton Res., 17(6), 1245–1271.

Maier-Reimer, E., U. Mikolajewicz, and A. Winguth (1996), Future ocean uptake of CO2: Interaction between ocean circulation and biology,Clim. Dyn., 12(10), 711–722, doi:10.1007/s003820050138.

Maiti, K., M. A. Charette, K. O. Buesseler, and M. Kahru (2013), An inverse relationship between production and export efficiency in theSouthern Ocean, Geophys. Res. Lett., 40, 1557–1561, doi:10.1002/grl.50219.

Martin, J. H., G. A. Knauer, D. M. Karl, and W. W. Broenkow (1987), VERTEX: Carbon cycling in the northeast Pacific, Deep-Sea Res., Part A, 34(2),267–285, doi:10.1016/0198-0149(87)90086-0.

Martin, P., R. S. Lampitt, M. Jane Perry, R. Sanders, C. Lee, and E. D’Asaro (2011), Export and mesopelagic particle flux during a North Atlanticspring diatom bloom, Deep Sea Res., Part I, 58(4), 338–349, doi:10.1016/j.dsr.2011.01.006.

Mcdonnell, A. M. P., and K. O. Buesseler (2010), Variability in the average sinking velocity of marine particles, Limnol. Oceanogr., 55(5),2085–2096, doi:10.4319/lo.2010.55.5.2085.

McNown, J., and J. Malaika (1950), Effects of particle shape on settling velocity at low Reynolds numbers, Eos Trans. AGU, 31, 74–82.Moore, J. K., and T. A. Villareal (1996), Size-ascent rate relationships in positively buoyant marine diatoms, Limnol. Oceanogr., 41(7),

1514–1520, doi:10.4319/lo.1996.41.7.1514.Moore, T. S., M. D. Dowell, and B. A. Franz (2012), Detection of coccolithophore blooms in ocean color satellite imagery: A generalized

approach for use with multiple sensors, Remote Sens. Environ., 117, 249–263, doi:10.1016/j.rse.2011.10.001.



http://dx.doi.org/10.5194/bg-6-85-2009

http://dx.doi.org/10.1029/2001GB001722

http://dx.doi.org/10.1038/nature13123


http://dx.doi.org/10.1371/journal.pone.0047887

http://dx.doi.org/10.1029/2011GL046735

http://dx.doi.org/10.1029/2011GB004099

http://dx.doi.org/10.4319/lo.2012.57.6.1591

http://dx.doi.org/10.1029/93GB01731

http://dx.doi.org/10.1029/2009GL041521

http://dx.doi.org/10.1029/2009GL041521



http://dx.doi.org/10.1002/2014GB004919

http://dx.doi.org/10.1016/j.dsr.2015.02.001

http://dx.doi.org/10.1029/2001GB001765

http://dx.doi.org/10.1016/j.rse.2013.09.009



http://dx.doi.org/10.1029/2010GB003868

http://dx.doi.org/10.1029/2010GB003868

http://dx.doi.org/10.1007/BF00355486

http://dx.doi.org/10.3354/meps221047

http://dx.doi.org/10.1002/2014GL061678

http://dx.doi.org/10.1016/j.atmosenv.2013.06.044



http://dx.doi.org/10.1007/s003820050138

http://dx.doi.org/10.1002/grl.50219

http://dx.doi.org/10.1016/0198-0149(87)90086-0

http://dx.doi.org/10.1016/j.dsr.2011.01.006

http://dx.doi.org/10.4319/lo.2010.55.5.2085

http://dx.doi.org/10.4319/lo.1996.41.7.1514

http://dx.doi.org/10.1016/j.rse.2011.10.001

Nowald, N., G. Fischer, V. Ratmeyer, M. Iversen, C. Reuter, and G. Wefer (2009), In-situ sinking speed measurements of marine snowaggregates acquired with a settling chamber mounted to the Cherokee ROV, in Ocean. 2009-Europe, pp. 1–6, IEEE, Bremen, doi:10.1109/OCEANSE.2009.5278186.

Passow, U., and C. L. De La Rocha (2006), Accumulation of mineral ballast on organic aggregates, Global Biogeochem. Cycles, 20, GB1013,doi:10.1029/2005GB002579.

Pilskaln, C. H., C. Lehmann, J. B. Paduan, and M. W. Silver (1998), Spatial and temporal dynamics in marine aggregate abundance, sinking rateand flux: Monterey Bay, central California, Deep Sea Res., Part II, 45(8-9), 1803–1837.

Ploug, H., M. H. Iversen, and G. Fischer (2008), Ballast, sinking velocity, and apparent diffusivity within marine snow and zooplankton fecalpellets: Implications for substrate turnover by attached bacteria, Limnol. Oceanogr., 53(5), 1878–1886.

Poulton, A. J., S. C. Painter, J. R. Young, N. R. Bates, B. Bowler, D. Drapeau, E. Lyczsckowski, and W. M. Balch (2013), The 2008 Emiliania huxleyibloom along the Patagonian Shelf: Ecology, biogeochemistry, and cellular calcification, Global Biogeochem. Cycles, 27, 1–11, doi:10.1002/2013GB004641.

Puigcorbé, V., C. R. Benitez-Nelson, P. Masqué, E. Verdeny, A. E. White, B. N. Popp, F. G. Prahl, and P. J. Lam (2015), Small phytoplankton drivehigh summertime carbon and nutrient export in the Gulf of California and eastern tropical North Pacific, Global Biogeochem. Cycles, 29,397–415, doi:10.1002/2014GB004979.

Riebesell, U., et al. (2013), Technical note: A mobile sea-going mesocosm system—New opportunities for ocean change research,Biogeosciences, 10(3), 1835–1847, doi:10.5194/bg-10-1835-2013.

Sanders, R., et al. (2014), The biological carbon pump in the North Atlantic, Prog. Oceanogr., 129, 200–218, doi:10.1016/j.pocean.2014.05.005.Sarmiento, J. L., and N. Gruber (2006), Ocean Biogeochemical Dynamics, Princeton Univ. Press, Princeton.Schmidt, K., C. L. De La Rocha, M. Gallinari, and G. Cortese (2014), Not all calcite ballast is created equal: Differing effects of foraminiferan and

coccolith calcite on the formation and sinking of aggregates, Biogeosciences, 11, 135–145, doi:10.5194/bg-11-135-2014.Schmittner, A., A. Oschlies, H. D. Matthews, and E. D. Galbraith (2008), Future changes in climate, ocean circulation, ecosystems, and

biogeochemical cycling simulated for a business-as-usual CO2 emission scenario until year 4000 AD, Global Biogeochem. Cycles, 22,GB1013, doi:10.1029/2007GB002953.

Schulz, K. G., et al. (2013), Temporal biomass dynamics of an Arctic plankton bloom in response to increasing levels of atmospheric carbondioxide, Biogeosciences, 10, 161–180, doi:10.5194/bg-10-161-2013.

Sharqawy, M. H., J. H. Lienhard, and S. M. Zubair (2010), Thermophysical properties of seawater: A review of existing correlations and data,Desalin. Water Treat., 16, 354–380.

Siegel, D. A., K. O. Buesseler, S. C. Doney, S. F. Sailley, M. J. Behrenfeld, and P. W. Boyd (2014), Global assessment of ocean carbon export bycombining satellite observations and food-web models, Global Biogeochem. Cycles, 28, 181–196, doi:10.1002/2013GB004743.

Smayda, T. J. (1970), The suspension and sinking of phytoplankton in the sea, in Oceanography and Marine Biology: An Annual Review,pp. 353–414, Taylor and Francis, London.

Smetacek, V., et al. (2012), Deep carbon export from a Southern Ocean iron-fertilized diatom bloom, Nature, 487(7407), 313–319, doi:10.1038/nature11229.

Steinberg, D. K., B. A. S. Van Mooy, K. O. Buesseler, P. W. Boyd, T. Kobari, and D. M. Karl (2008), Bacterial vs. zooplankton control of sinkingparticle flux in the ocean’s twilight zone, Limnol. Oceanogr., 53(4), 1327–1338, doi:10.4319/lo.2008.53.4.1327.

Stokes, G. G. (1850), On the effect of the internal friction of fluids on the motion of pendulums, Trans. Cambridge Philos. Soc., 9, 1–86.Tesche, M., P. Glantz, C. Johansson, M. Norman, A. Hiebsch, A. Ansmann, D. Althausen, R. Engelmann, and P. Seifert (2012), Volcanic ash over

Scandinavia originating from the Grímsvötn eruptions in May 2011, J. Geophys. Res., 117, D09201, doi:10.1029/2011JD017090.Turner, J. T. (2002), Zooplankton fecal pellets, marine snow and sinking phytoplankton blooms, Aquat. Microb. Ecol., 27, 57–102.Volk, T., and M. I. Hoffert (1985), Ocean carbon pumps: Analysis of relative strengths and efficiencies in ocean-driven atmospheric CO2

changes, in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, vol. 32, edited by E. T. Sundquist andW. S. Broecker, pp. 99–110, AGU, Washington, D. C.

Waite, A. M., Ö. Gustafsson, O. Lindahl, and P. Tiselius (2005), Linking ecosystem dynamics and biogeochemistry: Sinking fractionation oforganic carbon in a Swedish fjord, Limnol. Oceanogr., 50(2), 658–671, doi:10.4319/lo.2005.50.2.0658.

Welschmeyer, N. (1994), Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and pheopigments, Limnol. Oceanogr., 39(8),1985–1992.

Wilson, J. D., S. Barker, and A. Ridgwell (2012), Assessment of the spatial variability in particulate organic matter and mineral sinking fluxes inthe ocean interior: Implications for the ballast hypothesis, Global Biogeochem. Cycles, 26, GB4011, doi:10.1029/2012GB004398.

Xue, J., and R. A. Armstrong (2009), An improved “benchmark” method for estimating particle settling velocities from time-series sedimenttrap fluxes, Deep Sea Res., Part II, 56, 1479–1486, doi:10.1016/j.dsr2.2008.11.033.

Yasumoto, K., et al. (2014), Biogenic polyamines capture CO2 and accelerate extracellular bacterial CaCO3 formation, Mar. Biotechnol., 16,465–474, doi:10.1007/s10126-014-9566-z.



http://dx.doi.org/10.1109/OCEANSE.2009.5278186

http://dx.doi.org/10.1109/OCEANSE.2009.5278186

http://dx.doi.org/10.1029/2005GB002579

http://dx.doi.org/10.1002/2013GB004641

http://dx.doi.org/10.1002/2013GB004641

http://dx.doi.org/10.1002/2014GB004979

http://dx.doi.org/10.5194/bg-10-1835-2013


http://dx.doi.org/10.5194/bg-11-135-2014

http://dx.doi.org/10.1029/2007GB002953

http://dx.doi.org/10.5194/bg-10-161-2013

http://dx.doi.org/10.1002/2013GB004743



http://dx.doi.org/10.4319/lo.2008.53.4.1327

http://dx.doi.org/10.1029/2011JD017090

http://dx.doi.org/10.4319/lo.2005.50.2.0658

http://dx.doi.org/10.1029/2012GB004398


http://dx.doi.org/10.1007/s10126-014-9566-z

influence of plankton community structure on the sinking ... · the enclosed plankton community...

Documents